ImagePipe.Transform.Detector.ImageVision.Objects (image_pipe v0.1.0)

Copy Markdown View Source

ImagePipe.Transform.Detector for general objects, backed by the optional image_vision dependency (Image.Detection, RT-DETR, COCO-80). The dependency is not declared by ImagePipe — hosts opt in. When absent, available?/1 is false and callers fall back gracefully.

Objects with a confidence score below 0.5 are dropped before ImagePipe sees them.

Class names use the URL-facing underscore spelling (traffic_light); the model emits spaces ("traffic light"), which this adapter normalizes on both sides.